Home/Methodology

Transparency

How Slate calculates AI visibility

Slate measures AI visibility for software products in source-backed category prompts.

Methodology v1perplexitychatgptgeminiclaude

Visibility score

Visibility combines mention rate with answer position, using only eligible non-self-cited runs.

mention_rate = runs_mentioning / eligible_runs
position_score = mean(1 / sqrt(position_in_answer))
visibility_score = 100 x mention_rate x normalized_position_weight
rank = dense_rank by visibility_score within category

Categories tracked

Published categories with approved rosters and promoted ranking snapshots.

Prompt selection

Prompts are category-scoped and versioned with active prompt packs.

Refresh cadence

Local test snapshots are published manually; production cadence is weekly.

Product detection

Mentions are attributed to exact product aliases first, then brand/product context when unambiguous.

Brand vs product attribution

Leaderboards rank products, while brand pages roll up the best product performance.

Citations

Citation domains are normalized and stored separately from raw provider answers.

Confidence

Confidence reflects eligible run coverage and extraction completeness.

Sentiment

Sentiment fields are reserved for source-backed extraction and omitted when unavailable.

Limitations

Scores are a directional benchmark, not a guarantee of buyer preference or market share.

Manual review

Human review is used for ambiguous categories, candidates, claims, corrections, and snapshot quality gates.

Self-citation control

Slate-owned citations are retained for audit and excluded from ranking impact.